‘Diagramology’

Types of Feynman Diagram

Tim Evans
(2nd January 2018)

1. Pieces of Diagrams
Feynman diagrams1 have four types of element:-

• Internal Vertices represented by a dot with some legs coming out. Each type of
vertex represents one term in Hint .

• Internal Edges (or legs) represented by lines between two internal vertices. Each line
represents a non-zero contraction of two fields, a Feynman propagator.

• External Vertices. An external vertex represents a coordinate/momentum variable

which is not integrated out. Whatever the diagram represents (matrix element M,
Green function G or sometimes some other mathematical object) the diagram is a
function of the values linked to external vertices. Sometimes external vertices are
represented by a dot or a cross. However often no explicit notation is used for an
external vertex so an edge ending in space and not at a vertex with one or more other
legs will normally represent an external vertex.

• External Legs are ones which have at least one end at an external vertex. Note that
external vertices may not have an explicit symbol so these are then legs which just end
in the middle of no where. Still represents a Feynman propagator.

2. Subdiagrams
A subdiagram is a subset of vertices V and all the edges between them. You may also
want to include the edges connected at one end to a vertex in our chosen subset, V, but
at the other end connected to another vertex or an external leg. Sometimes these edges
connecting the subdiagram to the rest of the diagram are not included, in which case we
have “amputated the legs” and we are working with a “truncated diagram”.

3. Connection
We often talk about a vertex or a propagator, a line, being connected to other parts of the
diagram. What we mean is that we can find a path from one element to another without a
jump. You ignore the nature of the lines and propagators, ignore any arrows on propagators,
when you look for paths.
1
Feynman Diagrams are particular examples of what are called graphs in discrete mathematics. Graphs
are a set of vertices and a set of edges where an edge is a pair of vertices. A lot of the terminology of
Feynman diagrams comes from Graph Theory.
QFT Feynman Diagram Types, 2nd January 2018 2

4. Component
A component is the maximal set of connected elements, all the elements which have paths
running between all the others in the component. Each component will be disconnected
from any other components.
For example in Scalar Yukawa theory we get the following diagram in ψψ → ψψ scattering

p1 q1

p2 q2
(1)
This has diagram has three components.

5. Vacuum Diagrams
A vacuum diagram is a diagram which is not connected to any external leg. An external
leg of a diagram is a propagator carrying at least one of the arguments of the Green function,
coordinates {yi } or {zf } or momenta {pi } or {qf } in our conventions. Therefore is not a
of any coordinates, it is just a constant. These diagrams are disconnected components and
they are never connected to any external leg.
Vacuum diagrams capture the contribution of virtual fluctuations to the vacuum of a
fully interacting theory. That is they are contributions to h0|S| 0i.
The following is a vacuum diagram in λφ4 theory

(2)
The top component of the diagram in (1) is also an example of a vacuum diagram in Scalar
Yukawa theory.

6. Tadpole Diagrams
Diagrams with a tadpole contribution are contributions to the vacuum expectation value
of a field, hφ̂i. A tadpole is a part of a diagram which can be disconnected from all external
legs if you cut just one line. That is a tadpole looks like a blob with one leg — the tail of the
tadpole — coming out. These are contributions to expectation values of fields, h0|φ(x, t)| 0i
and so forth.
In practical work you should choose to work with fields where tadpole diagrams are zero.
That means that you work with fields with zero expectation values, if necessary working
with new fields η̂j defined by η̂j (x) = φ̂j (x) − h0|φ̂j (x)| 0i. While not necessary, this greatly
simplifies calculations. See the Unification course as this is needed when we have symmetry
breaking. However in our course we will not enforce this though for most of our examples
the tadpole diagrams are zero.
QFT Feynman Diagram Types, 2nd January 2018 3

7. Connected Diagrams
Diagrams with just one component are called connected diagrams. They have no discon-
nected vacuum components so they contribute to what are called connected Green functions
Gc where the vacuum expectation values are properly normalised:-
1
Gc (x1 , . . . , xn ) = h0|T φ(x1 ) . . . φ(xn )S| 0i = hΩ|T φ(x1 ) . . . φ(xn )S| Ωi . (3)
h0|S| 0i

The diagrammatic notation for an n-point connected Green function, Gc (x1 , . . . , xn ), is

a circle with diagonal hatching with n legs emerging:-

x1
x2
Gc (x1 , . . . , xn ) =

xn x3
(4)

Note for generic green functions G(x1 , . . . , xn ) = h0|T φ(x1 ) . . . φ(xn )S| 0i which may contain
disconnected parts, the diagrammatic notation is the same except that there is no hatching
in the circle.
The two-point connected diagrams contribute to what is called the full propagator
Π(x − y) where
Π(x − y) = hΩ|Tφ(x)φ(y)S| Ωi . (5)

8. One-Particle Irreducible — 1PI

A One-Particle Irreducible diagram — 1PI diagram — is a diagram which does not fall
into two pieces if you cut one internal line. By convention you only include the propaga-
tors/lines connecting two vertices within the 1PI diagram, you never include the propaga-
tors/lines connected to one vertex on the edge of the 1PI diagram and then to something
outside the diagram. We say we truncate or amputate the (external) legs of these dia-
grams.
Consider the following example to illustrate the 1PI concept.

(6)

Here the external legs have been amputated/truncated so they are not present to be cut.
However the line in the middle if cut, as indicated by the dashed line, would leave this
diagram in two pieces, so the whole diagram is not 1PI. However the two subdiagrams
on either side of the dashed line, if the connecting propagator has been removed (the one
QFT Feynman Diagram Types, 2nd January 2018 4

intersecting the dashed line), are both in fact 1PI subdiagrams in this case, both two-point
diagrams (though special cases of this) as each subdiagram needs to be connected up to lines
from the vertices on the edge to make a proper Feynman diagram.
The diagrammatic notation for an n-point 1PI function, Γ(x1 , . . . , xn ) which is a sum of
all n-point 1PI diagrams, is a circle with two orthogonal diagonal hatching and n vertiex
blobs (or sometimes a stub of a line) to indicate that no propagators are included for these
external parts:-
x1
Γ(x1 , . . . , xn ) = x2
xn x3
(7)

9. Self-Energy Σ
Self-energy contributions are parts of diagrams which represent quantum corrections to a
propagator. Formally they are two-point 1PI one particle irreducible diagrams. These are
diagrams which can be dropped on top of one type of line to give a new legal diagram. The
function Σ(p) is called the self-energy. It is the two-point 1PI (one-particle irreducible)
function and it is described by the sum of 1PI diagrams with two amputated legs. That is the
two external legs which could be added to give a legal diagram but the propagators usually
associated with these legs is not part of this contribution to Σ by definition. A 1PI diagram
is one which can not be cut into two separate parts by cutting one line. In particular this
means there are no contributions from external lines on a 1PI diagram (I draw them as little
stubs not long lines).
For instance a diagram contributing to the full propagator Π(x − y) of (5) may be
expanded in terms of the free propagator ∆(p) in powers of the self-energy Σ as follows

Π(p) = ∆(p) + ∆(p)Σ(p)∆(p) + . . . (8)

The second term may be cut into two by cutting either of the lines representing the ∆(p)
factors, which are external lines in the diagram. The remaining part will contribute to Σ is
that can not itself be cut into two pieces. If it can then we are dealing with a contribution
of the form ∆(p)Σ(p)∆(p)Σ(p)∆(p) or perhaps one of the other higher order terms.
Writing this out fully we find a very useful result for the full propagator Π(x − y) of (5),
namely
∞
X
Π(p) = ∆(p) (Σ(p)∆(p))n (9)
n=0
1 i
Π(p) = = . (10)
(∆(p))−1 − Σ(p) p2 − m2− iΣ(p)

That is the pole of the full propagator is at p2 = m2 + iΣ(p2 = m2 ) = m2phys , a self-consistent

equation for the true physical mass m2phys . Should iΣ(p2 = m2 ) be real this means that the
physical mass is not m but at the position of the pole in the full propagator. The quantum
fluctuations encoded in Σ have interacted with our particle, altering the way the particle
propagates as compared to the free (no-interaction) propagator ∆.
QFT Feynman Diagram Types, 2nd January 2018 5

For example in λφ4 theory, one of the self-energy contributions is the following two-point
1PI diagram

(11)
Note no external legs/propagators are included on this diagram.
See (6) for a further example.

10. Momentum values and On-Shell Green functions

If you have an N -point Green function G({x}) it is just a function of N coordinates in
space-time. We can find the Fourier transform of this in the usual way, it is a function
of N four-momenta. To do this we will need an integration over the full four-dimensional
space-time for each of the N coordinates, and the inverse will have N four-dimensional
momentum integrations. Note that this means that in principle the N energy and three-
momenta arguments can take any value.
In practice we know the physics will lead to an overall energy-momentum conservation
delta function for each component of any diagrammatic contribution so the functions are
zero when the N energy-momenta variablestogether do not satisfy such general conservation
principles.
However a Green function in momentum space does not have to have it’s arguments (the
four-momenta of each external leg) what is called on mass-shell. That is each of the four-
momentum arguments need not satisfy some p2 = m2 relation. These external four-momenta
need not be the energy and three-momentum of a physical particle. For instance this Green
function might be used as part of the description of some larger physical process and we
need to know the contribution coming from virtual particles within this process. However it
is mostly a matter of mathematical convenience that we work with these generalised Green
functions with unrestricted four-momenta values.
However when we want to use a Green function to calculate a matrix element, then we
do need real physical particles (at plus/minus infinity in time). The part of a Green function
which represents this matrix element must then have the external legs on mass-shell. This
is a particular limit of a Green function, most values of a Green function are for unphysical2 ,
virtual particles. You will note that the relationship between a matrix element and a Green
function essentially removes the propagator for the external legs and puts their (external)
momenta on-shell to the values specified in the matrix element. That is a matrix element
takes a particular part of a general Green function and does not access the whole information
contained in a general Green function.

2
As just noted, off-shell Green functions can contribute to a larger physical process so in some sense they
still contain real useful (i.e. physical) information just a little more indirectly.